Post on 28-Jan-2023
Reconstructing Proto-Indo-European Deponents
Laura Grestenberger
Lgresten@fas.harvard.edu
33rd East Coast Indo-European Conference
6-8 June 2014, Virginia Tech
1 Introduction
� What kind of transitive constructions were marked with middle morphology in Proto-Indo-
European?
(Late) PIE had a bivalent voice system, with active vs. middle (�non-active�) endings (Jasano� 2003,
Weiss 2009, Fortson 2010).
(1) PIE active/middle endings
active middle
primary secondary primary secondary
1sg. *-mi *-m *-h2e-r *-h2e
2sg. *-si *-s *-th2e-r *-th2e
3sg. *-ti *-t *-o-r ; *-to-r *-o; *-to
1pl. *-me- *-me(-) *-medhh2(-r?) *-medhh2
2pl. *-te- *-te(-) *-dh(u)u“e(-r?) *-dh(u)u
“e
3pl. *-(é)nti *-(é)nt *-ro(-r?); *-nto-r *-ro; *-nto
1. General question: What was the distribution of the PIE active and middle endings?
2. Speci�c question: What kind of transitive, formally middle verbs did PIE have?
3. Even more speci�c question: Can we reconstruct voice mismatches for PIE, in particular syn-
tactically active, transitive verbs with middle morphology (deponents, narrow de�nition)
Illustration: How do we predict the distribution of active vs. middle morphology in the following
transitive clauses?
1
(2) a. Xnom Yacc gu“hénti �X slays Y�
b. Xnom Yacc u“éstor �X wears Y�
c. Xnom Yacc péh2s(t)or �X protects Y�
2 Background: middles in Indo-European
Non-active morphology is found in the same syntactic environments cross-linguistically (Geniu²iene
1987, Klaiman 1991, Kemmer 1993, 1994, Embick 1998, Kaufmann 2007, Kallulli 2007, 2013, Alexi-
adou and Doron 2012, Alexiadou 2013, etc.) = canonical middles.
� Most of these environments are intransitive (inchoatives, anticausatives, re�exives, statives, etc.).
� Middle/non-active morphology in intransitive environments (cp. Gonda 1960, Holli�eld 1977)
� Anticausatives/inchoatives:
� �Caland system� verbs, e.g., Ved. práthate `spreads out', pávate `becomes clean', rócate
`shines', Gk. τρεφομαι `grows up', μελδομαι `becomes soft', θερομαι `becomes hot', etc.
(Rau 2009, 2013)
� Re�exives & reciprocals:
� Vedic: ±umbháti `adorns' � ±úmbhate `adorns oneself', recipr. spárdhante `they �ght'
� Greek: λουω `wash' � λουομαι `wash myself', recipr. μαχονται `they �ght'
� Tocharian: A ytäs˙tär `adorns oneself'
� Hittite: zahˇhˇanda `they �ght',
� Modern Greek: pleno `wash' � plenome `wash myself', recipr. adelfo-skotonomaste
`we brother-kill each other'
� (Medio)passives:
� Latin amo � amor , etc.
� Vedic ±r˙n˙vé `is heard'/`can be heard', dadr
˙±é `is visible/is seen', bruvé `is spoken', etc.
� OIr.: ber(a)ir `is (being) carried', benair `is (being) struck', melair `is (being) ground',
etc.
� Statives:
� Ved. ±áye `lies', Ved. ´a mahe `is able', Gk. κειμαι, Hitt. kitta(ri) `lies', dukkari `is
visible', etc.
� Verbs of motion:
� Vedic: plávate `swims, �oats', g´ahate `enters, immerses', násate `returns'
� Greek: πετομαι `�y', ερχομαι `come'
� Latin: vagor `wander', gradior `advance', orior `rise'
� Modern Greek: erhome `come', a�knume `arrive'
� ... but Indo-European also has transitive structures that combine accusative objects with non-
active morphology.
2
2.1 Self-benefactive/�a�ected experiencer� constructions
� Vedic: yájati `sacri�ces' � yájate `sacri�ces sth. for one's own bene�t', bhárati `carries' �
bhárate `takes/carries sth. for oneself', pácati `cooks' � pácate `cooks for oneself', etc.
� Greek: φερω `carry' � φερομαι `carry for myself, win', ἀειρω `lift, raise' � ἀειρομαι `lift up for
myself', τιθημι `place, put' � τιθεμαι `place for myself', etc.
Examples:
(3) Vedic1
a. RV 8.45.39:
´aprvb
tayour
et´athese.acc
vacoyújaspeech.yoked.acc
hár�fallow.acc
gr˙bhn
˙e
seize.1sg.pres.midsumádrathawith.chariot.acc
�I hold onto these two fallow bays of yours, yoked by speech, along with their chariot.� (=
`seize for myself', vs. gr˙bhn´ami)
b. RV 1.161.6:
b¯˙haspátir
Br˙haspati.nom
vi±várupamall.forms.acc
úpajataprvb.drive.3sg.ipf.mid
�Br˙haspati drove near (the cow) of all forms.� (= `drove for himself', vs. úpajat)
(4) Greek, Il.23.413:
η
surelyτ'part
αν
part
ἐγω`
Iτα`
def
πρωτα
�rstλαβω`ν
seize.ptcp.nomκλισvιην
shed.accδε`
part
φεροιμην
carry.1sg.opt.mid
�Surely I will seize the �rst prize and carry it to my shed (for myself)�
2.2 Experiencer verbs
= psych verbs, �cognition middles� (Kemmer 1993)
(5) Transitive experiencer verbs in Vedic and Greek
Non-active verb Case on theme
Vedic chandayate `likes, takes pleasure in' (√chad) acc; loc
jus˙ánta `they taste, enjoy' (√jus
˙) acc (gen)
pánanta `they admire' (√pan) acc
mr˙s˙yate `forgets' (√mr
˙s˙) acc
Greek αζομαι `am in awe of, fear' acc
αγαμαι `marvel, be angry at' acc, dat
βουλομαι `wish, will' acc, inf.
1Translations are from Jamison and Brereton (2014).
3
(6) a. RV 7.18.21: ná te bhojásya sakhyám mr˙s˙anta �they did not neglect their partnership with
you, who provided for (them).�
b. Od.2.67: ἀγασσvαμενοι κακα` εργα �angry at your evil deeds�
(Cp. transitive experiencer verbs in Modern Greek (Zombolou and Alexiadou 2014): esthanome `feel',
osmizome `smell', fovame `fear', gevome `taste', kapsurevome `fall in love with', drepome `be ashamed
of', sevome `respect', etc.)
2.3 (Some) re�exive constructions
Some re�exive/body part constructions are also formally transitive (with an accusative re�exive pro-
noun or anaphor):
(7) Vedic:
a. RV 10.8.3:
árus˙�r
reddish.nom.pl.f... r
˙tásyatruth.gen
yónauwomb.loc
tanvòself.acc.pl
jus˙anta
enjoy.3pl.aor.mid
�The reddish [�ames] (...) enjoy themselves in the womb of truth.�2
b. 1.147.4d:
ánuprvb
mr˙ks
˙�s˙t˙a
injure.3sg.aor.midtanvàmself.acc
duruktaíh˙bad.words.instr
�He shall injure himself with (his) evil words.�3
2.4 Statives and verbs of motion
Re�exes of *u“ésto `wears' are transitive:
(8) RV 4.2.19:
r˙támtruth.acc
avasrannwear.3pl.ipf.mid
us˙áso
dawn.nom.plvibhat�h
˙radiant.nom.pl
�The radiant dawns clothed themselves in truth�4
(9) Hitt., ABoT 4 + I 24f.:
we²²anda=mawear.3pl.pres.mid=part
i²hˇharwantu²
blood.red.acc.plTÚG-Hˇ I.A-u²clothes-pl-acc.pl
putaliyante²=agirded.pres.ptcp.nom.pl=and
`They are wearing blood-red clothes and are girded up.�
2Jamison and Brereton (2014): �... the ruddy females [=�ames? dawns?] (...) �nd pleasure in their own bodies withinthe womb of truth.�, Geldner (1951): �(...) haben die rötlichen (Flammen) ... an der Wiege der Opferordnung an sich selbstWohlgefallen gefunden.�
3Jamison and Brereton (2014): �he should bring harm upon his own body by his evil words.�4Insler (1968).
4
... as are the re�exes of *séku“ eto `joins, accompanies' in some languages:
(10) a. Ved., RV 4.7.11c:
v´atasyawind.gen
med˙ím
˙roar.accsácatefollow.3sg.pres.mid
(...)
�He follows the roar of the wind�
b. OIr., Ml. 19b11:
níneg
sechetarfollow.3pl.pres.non-act
immurguhowever
ordorder.acc
ocat
suidiuthis.dat
�They do not, however, follow the order in this�
c. Lat., Plaut., Aulularia 4.7.16:
I,go.ipv
iamat.once
sequorfollow.1sg.pres.pass
te,you.acc
matermother
�Go! I (will) follow you at once, mother.�
2.5 Deponents
Finally, there are transitive middles that syntactically behave like formally active agentive verbs and
which usually have formally active synonyms or near-synonyms. There is no obvious synchronic
motivation for middle morphology on these verbs.
(11) Active/deponent synonyms
Language Deponent verb Active verb Meaning
Latin hortor moneo `encourage, incite
tueor servo, salvo `keep safe, watch over, protect'
furor clepo, rapio `steal, rob'
Hom. Greek ἐρυομαι φυλασσω `protect, guard'
ειρομαι ἐρωταω `ask, question'
δαιομαι σχιζω `divide, separate'
Modern Greek skar�zome epinoo `contrive, devise'
(meta)hirizome hrizimopio `use'
katarieme anathematizo `curse'
(This is a relatively small class of verbs: Zombolou and Alexiadou (2014) claim that only 11% of
their corpus of Modern Greek media tantum are agentive and transitive and do not �t into any other
established category of uses of middle morphology.)
Syntactically, these behave like formally active transitive verbs:
5
� They take accusative objects:
(12) Greek, Il. 3.278-9:
και`
andοι
who.nom.plὐπενερθε
beneathκαμοντας
passed.on.ptcp.acc.plανθρωπους
men.accτ΄ινυσθον
punish.2du.pres.mid.
�and (you) who in the underworld punish the men who have passed on�
(13) Vedic, RV 2.23.4a-b:
tr´aya-seprotect-2sg.pres.mid
jánam˙man.acc
yáswho.nom
túbhyam˙you.datd´a±atworship.3sg.subj.act
�.. you rescue the man who will do (ritual) service for you.� (Jamison and Brereton 2014)
� They passivize i� distinct passive morphology is available
(14) Vedic deponent passives
Root Deponent Passive
�d˙
´�t˙-t˙e `praises' �d
˙-yá-te `is being praised'
idh ind-dhé `kindles' idh-yá-te `is being kindled'
rabh rábha-te `seizes' rabh-yá-te `is being seized'
(15) Greek, deponent δωρεομαι `give, endow with': Herodotus, Hist . 8.85.3:
Φυλακος
Phulakos.nomδε`
part
ἐυεργετης
benefactor.nomβασιλεος
king.genἀνεγραφη
down.write.aor.pass.3sgκαι`
andχωρηι
land.datἐδωρηθη
endow.aor.pass.3sgπολληι
much.dat
�Phulakos was recorded as benefactor of the king and endowed with much land.�
� They make agent nouns
(16) Vedic deponent agent nouns
a. tra-tár- `protector' (tra `protect')
b. �d˙i-tár- `praiser, worshipper' (�d
˙`praise')
c. ks˙at-tár- `server' (ks
˙ad `serve, prepare')
6
(17) Greek deponent agent nouns
a. ῥυτηρ `protector' (ῥ῀υμαι `protect, guard')
b. ληιστηρ `robber' (ληιζομαι `rob, plunder')
c. λωβητηρ `slanderer' (λωβαομαι `slander, mistreat')
Conclusion: Languages with an active/non-active voice system have morphologically non-active
transitive verbs that cannot be synchronically motivated = deponents.
(18) De�nition of deponency (Grestenberger 2014):
�In an active�non-active voice system, a deponent is a syntactically active verb whose surface
subject is an agent and whose �nite forms are morphologically non-active.�
� Can we reconstruct such verbs for PIE?
... in principle, yes: PIE had the same type of voice system.
3 Proto-Indo-European deponents
Criteria for determining PIE deponents:
� Synchronic deponent status in at least two separate branches: �Two separate branches�
means non-adjacent language families without shared innovations.
� No synchronic motivation for non-active morphology: If there is some synchronic mo-
tivation for non-active morphology on an agentive verb in one branch, like analogy with a
semantically or morphonologically similar verb, its claim to PIE deponent status is weakened,
unless there is corroborating evidence from other branches in which there is no synchronic mo-
tivation for non-active morphology. In other words, the non-active morphology on such verbs
must be non-trivial from a synchronic point of view.
This is especially relevant for verbs of speech and certain verbs of visual perception, which vary
greatly across the individual branches.
� Morphological correspondence: If cognate deponents in di�erent branches correspond in
their derivational morphology, they are more likely inherited than if they have di�ering verbal
morphology.
� Syntactic correspondence: Identical object case and inherited agent nouns or verbal adjec-
tives with the same syntactic behavior may provide additional evidence for deponent status.
7
3.1 *deh2-i“- `divide, distribute'
� Evidence: Ved. dáyate `distributes', Gk. δαιομαι `distributes' (cp. Δαιτωρ, δαις, -τος `portion').
� Other forms (non-deponent): Root aorist: Ved. áva adat (MS), d�s˙va (YV), pres. dyáti `divides'
(AV).
� Reconstruction: (late) IE *dái“h2-e-toi
“5 ← *h2e-conjugation i -present 3sg. *déh2-i
“-e, 3pl. dh2-i
“-
énti (Jasano� 2003: 105�.)
3.2 *h1u“egu“
h `speak solemnly, praise'
� Evidence: Ved. 3pl. pres. óhate `praise', ptcp. óhana-,6 OAv. 1sg. aojoi , 3sg.ipf. aog@da, ptcp.
aojana-, Gk. 3sg.ipf. ευκτο (Thebaïs 3.3), remodelled as a thematic present: ευχομαι (Myc.
eu-ke-to /eukhetoi/).
� Other forms (non-deponent): Lat. voveo `vow' (< *h1u“ogu“
h-éi“e/o-), ?Ved. vaghat- `praiser' (if
from *h1u“o/egu“
h-n˚t-, EWA II: 539)), ?Arm. 2sg.ipv. gog `speak!' (aor. *gog(e)- < *u
“ogu“
h, Klin-
genschmitt (1982: 275)).
� Reconstruction: root shape *h1u“egu“
h seems to require a 3sg. *h1é-h1ugu“h-to(i“) (LIV2: 253, Vil-
lanueva Svensson 2012: 335), but a full grade middle present *h1eu“gu“
h-to(i“) would be more
regular.
3.3 *peh2-s `protect'
� Evidence: Hitt. pahˇ²a(ri) `protects',7 Toch. A (class II) 3sg.pres. pas
˙tär , 3pl. pasantär `protect'
(< *-s- or *-s“ke/o-present), Jasano� 1988: 230f., fn. 10, 2003: 136, 182f., 2012a) ∼= Toch. B
pas˙tär , 3pl. paskentär (< *-s“ke/o-present).
� Other forms (non-deponent): Ved. p´ati , OAv. pat˜`protects', subj. pasati (= s-aorist subjunctive,
LIV2: 460, Narten (1964: 168f.)); OCS paso `graze, herd', Lat. pasco `graze'; PPP pastus, pastor
`shepherd'.
� Reconstruction: s-present 3sg. *péh2-s-or .
3.4 *h2neh3 `scorn, reprimand'
� Evidence: Gk. ονομαι `scorn, reprimand' (2sg. ονοσαι, aorist ὠνοσvαμην), Hitt. hˇannari `contest
at law, sue; judge'; Toch. n˘ak : Toch. B class I subj. (inf) naktsi , pret. III n˘aksate, class VIII
5The root shape /day/ is unexpected in both languages; Jasano� 2003: 102�.: lack of compensatory lengthening in i -presents to laryngeal-�nal roots is due to the �AHIHA-rule�: *-AHIHA- > *-AIHA- (A = any vowel, I = i
“or u
“): *déh2-i-h2e
> *dé-i“-h2e), with subsequent analogical extension of the new root shape to contexts where compensatory lengthening would
regularly take place, like the 3sg. *deh2-i“-e. In Greek, the glide was preserved in analogy with form like δαινυμι `give a feast',
fut. δαισω, δαις, -τος `portion', etc.6Also ohaná-.7Hitt. pah
ˇ² - is not a mi -verb (pace LIV2: 460); active forms are practically non-existent until the Neo-Hittite period and
Kloekhorst (2008: 612) points out that pahˇ² - is �activized� as a h
ˇi - as a mi -verb when it is �nally transformed into an active
verb.
8
pres. n˘aks˙tär , Toch. A nakäs
˙tär `blames, reprimands' (Hackstein 1995: 65�., Malzahn 2010:
677f.).
� Other forms (non-deponent): OIr. -antar `is blemished' (Watkins 1962: 116�., Pinault (1982)
loc.cit., LIV2 loc.cit.).
� Reconstruction: Full grade I present *h2énh3-(t)or (or zero grade *h2n˚h3-(t)or?).8
3.5 *med `measure (out)'
� Evidence: Gk. μηδομαι `devise, contrive' (aor. ἐμησατο) and μεδομαι `take care of' (+gen.);
athem. Hsych. μηστο · βουλευσατο< present injunctive *med-to = OIr. �long-vowel preterite�
·mídair `judged' (Narten imperfect, Jasano� 2012c, cp. also LIV2: 423), OIr. midithir `judges'
(< *med-i“e/o-); OLith. pam§emi < *pa-m§edmi , inf. pam§edeti (also (pa)m§edyti) `imitate, ape'
and Latv. m�edît `imitate' < Proto-Baltic *med-mai `measure' (Villanueva Svensson 2006); OAv.
3sg. aor.subj. masata `shall measure out' (Hintze 2000), YAv. 3pl. pres. v�-maδaiian˙ta (V.7.38,
V.7.40), 3pl. pres.subj. v�-maδaiiån˙te; Lat. medeor `help, heal' (+dat., < *med-eh1-i
“e/o- (or
*med-ei“e/o-?).
� Other forms (non-deponent): Goth. mitan `measure' (OE metan, OHG mez(z)an, etc., cp. Feist
1939: 363f.) < *med-e/o-; Gk. μεδων, μεδεων `ruler, ruling' (Hom.+).
� Reconstruction: 3sg. *m´ed-or , 3pl.méd-ror `measure'9 (Villanueva Svensson 2006), but probably
rather a *h2e-present 3sg. *m´ed-e, 3pl. *méd-r˚(s),10 with subsequent generalization of di�erent
ablaut grades (R(e) in Greek, OIr., and Baltic; R(e) in Greek, OIr., and Germanic).8The Greek present could re�ect either *h2énh3-(t)or or *h2n
˚h3-(t)or more or less directly. Both *h2n
˚h3-C- and *h2enh3-c
would have given *ano-C- with subsequent assimilation to *ono-C- (thus Hackstein 1995: 66, who also dismisses Hom. ωνατο
as evidence for �nal *h2; see also Pinault (1982: 20�.)).The Hittite form is more problematic. While a full grade I middle *h2énh3-or should have given *HanH-V- > h
ˇann-V- (cp.
hˇarra- `crush' < *h2érh3-V-, Melchert 1994: 79f., Kloekhorst 2008: 300f.; on the loss of *h3 in other positions see Melchert1994: 72�.), *h3 was apparently also preserved in medial position in some cases, the circumstances of which are not clear, cp.Hitt. walh
ˇ- < *u(e)lh3
““-, Kloekhorst 2008: 945f.; Hitt. lah
ˇu- `pour' < lóh3-u
“-, Melchert 2011. If *(C)R
˚h3V gave *(C)aRh
ˇV , as
Kloekhorst, loc.cit. suggests, a zero grade middle *h2n˚h3-ór , on the other hand, should presumably have surfaced as *h
ˇanh
ˇari ,
and even if gemination took place, the resultant verb should behave like dukkari `is visible' (re�ecting accent on the ending).Since full grade seems to be the more expected ablaut grade of old media tantum and the medial re�exes of *h3 are contestedeither way, it therefore seems more prudent to operate with a full grade I middle *h2énh3-(t)or for Greek and Hittite.Tocharian requires a full grade II root shape *n˘a- < *h2neh3 which it extended with an element -k- (Hackstein 1995: 66f.).9The meaning `measure' is attested in Avestan, Germanic, and Italic (Lat. modus `measure, mode', Osc. meddíss `judge' <
`the one who shows (*dei““k) the measure (*med), the established mode'), `devise' in Greek, `heal' in Latin (presumably related
to `take care of' seen in Gk. μεδομαι), `judge' in Old Irish (and cp. Osc. meddíss), `rule' in Greek μεδων, and `imitate' inBaltic. LIV2: 423 essentially follows Benveniste (1969: 123�.), who argues that the meaning `measure out (to establish/decidesomething)' was the original meaning of this root. This is seen most clearly in Avestan and in the Italic nominal derivativesof this root. This then developed into `measure out, establish (the right measure) for somebody' > `take care of' in Greekand Latin. For Latin, this actually makes a denominal origin *med-eh1 (instr.) + *-i
“e/o from the instrumental of the root
noun underlying Osc. medíss the most likely derivation of medeor `help, heal'. This formation should have meant `have/bewith the right measure (for)', and the fact that medeor usually takes the dative (rather than the accusative) con�rms this.The development from `measure out' to `establish, rule, judge' (Gk. μεδων, OIr. midithir) on the one hand and `devise' (Gk.μηδομαι) on the other is also not too surprising. A parallel for Baltic `imitate' is the development of Skt. prati-ma- `imitate'from ma `measure' (Villanueva Svensson 2006: 97).
10To avoid setting up an ablauting middle paradigm for PIE (I am grateful to Jay Jasano� for this suggestion). *h2e-conjugation verbs originally had R(o/e)-ablaut, but there is no principled reason why Narten root *h2e-verbs should not havehad R(e/e)-ablaut. This Narten ablaut may have been associated with a particular Aktionsart in the present stem, see Kümmel(1998) and Melchert (To appear).
9
3.6 *i“i-i
“eh2- `demand, seek'
� Evidence: Gk. διζημαι `seek' ≈ stem of Vedic ´�mahe `we are asking, pleading', 1sg. iye, ptcp.
iyaná- (NB the accent), García Ramón 1993, 1999b, LIV2: 310f.: < *i“í-i“h2-; Toch. B subj. II
yas˙tär `will ask, beg' < *ih2-s“ke/o- (Malzahn 2010 Hackstein 1995: 184�.; 242).
� Other forms (non-deponent): Ved. y´ati `asks, requests', aor.subj. yasat `shall plead'; Ved. yatár -
`avenger' = Gk. Ζητηρ (Hsych. Ζητηρ · Ζευ`ς ἐν Κυπρῳ) < *i“eh2-tér - `seeker' (García Ramón
1999b).
� Reconstruction: *i“í-i“
@/ih2-to(i“) `seeks, requests'.11
4 Possible Proto-Indo-European deponents
4.1 *treH/*tra `protect'
� Evidence: Ved. tr´ayate `protects' = YAv. θraiien˙te `protect' (Yt.13.146, also inf. θraiioidiiai ,
Y.34.5, Y.11.9), Ved. (s-aor.) ipv. tr´asva, pl. tr´adhvam (= OAv. θrazdum, Y.34.7, Y.58.5), subj.
trasate (Narten 1964: 131f.), Ved. tratár - `protector' = Av. θratar .
� Reconstruction: (pre-)Proto-IIr. *tra-i“a-ta(i
“) `protects'.
4.2 *gres/gras `devour'
� Evidence: Ved. grásate `devours' (perf. mid. opt. jagras�ta, ptcp. jagrasaná-).
� Other forms (non-deponent): Gk. γραω `eat, gnaw' (Call., also Hsych. γρα · φαγε), Cypr. Gk.
ipv. ka-ra-si-ti /grásthi/ (4th century BCE, Masson 1983: 280).
� Reconstruction: *gras-e-toi“`devours'12
4.3 *u“er `protect; fend o�, stop'
� Evidence: Gk. ερυμαι `protect' (thematized ἐρυομαι), ῥυμαι (also ῥυομαι), perf. ειρυμαι.13
11Greek διζημαι introduced the full grade of the root (García Ramón 1993, za-/ze- < *i“eh2-, cp. ζητεω `seek' (Hom.+) and
the full grade verbal adjective Ved. yatá-, OAv. yata-).12The middle in�ection in Vedic is non-trivial, since other verbs of ingestion are formally active (átti `eats', a±n´ati `eats',
ághas `devoured', píbati `drinks', cp. Buck and Petersen 1945: 327�.). Since the same is more or less true in Greek (ἐσθιω `eat',εδω `eat', but fut. εδομαι, aor. εφαγον, βεβρωκα `have devoured', π΄ινω `drink' vs. πατεομαι `eat, taste', ἐρεπτομαι `feed on'), wecould posit a late Proto-Indo-European/pre-Graeco-Aryan deponent whose middle in�ection was given up in Greek.
13The lack of a digamma e�ect in Greek has given rise to comparison with Latin servo `save, protect'; this is rejected bySolmsen (1901: 245�.) who also discusses the digamma problem. LIV2: 685, n. 4 suggests that the full grade of ερυμαι<*u“ér-u-mai was introduced in analogy to the active or the subjunctive, but the root is solidly deponent in Greek and there
is independent evidence that media tantum paradigms had full grade (see Villanueva Svensson 2012). In light of this, theapparent ablaut suggested by the zero grade variant ῥυμαι < *wr-u-mai is more problematic.Hackstein (2002: 124f.) tries to solve the morphological and phonological problems in Greek by assuming that the underlying
root is the same as that of Gk. 3pl. ορονται `are watching over, taking care of', YAv. n� haraite `preserves', etc., for whichLIV2: 534 has a separate entry 1. *ser , and of Lat. servo `save, protect'. He posits a root *su
“erh3 (the meaning of which
is never de�ned) and argues that the di�erent root shapes seen in Greek, Latin, Avestan, etc., can be accounted for by theProto-Indo-European metathesis rule *u
“R˚(H) > *Ru(H) (cp. Mayrhofer 1986: 161f.). Besides the full grade *su
“erh3 and the
regular zero grade su“r˚h3, the metathesized weak stem variant *sruh3 would give Gk. ῥυ- (ῥυμαι) directly. While this solution
10
� Other forms (non-deponent): Ved. várutha- `protection', varutár -/várutr�- `protector' (besides
vartár - `defender', cp. EWA II: 512f.); maybe some forms of Ved. vr˙`cover' (if this continues
forms of *u“er `protect', *Hu
“er `lock in, keep safe' (Lat. aperio `open', etc.), and *u
“el `lock in,
cover up' (Gk. εἰλεω `hem in', etc.), e.g., subj. várate, pres. vr˙n˙óti (LIV2: 684f.).
� Reconstruction: (late) *u-present *u“ér-u-to `protect, defend, fend o�'?
5 Results
� 6 relatively secure forms:
� PIE i -present *deh2-i“-e ⇒ pre-Graceo-Aryan *dai
“etoi
“`distributes'
� (late) IE *h1éu“gu“
h-to(i“) (or *h1é-h1ugu“h-to(i
“)) `praises'
� PIE *péh2-s-or `protects' ⇒ *péh2“-s-tor (⇒ *peh2-s“ke/o-?)
� PIE *h2énh3-or `blames, scorns' ⇒ *h2énh3-tor/i“
� PIE *m´ed-e/*méd- `measures' ⇒ IE *med-to(i“), ⇒ *m˘ed-e-to(i
“)
� (late) IE *i“í-i“ih2-toi
“`seeks'
Middle morphology not easily explicable within the PIE voice system = deponents
� Full grade (R(e)) prevails in the older forms (NB full grade middles do not imply �Narten�
paradigms/roots, cp. Villanueva Svensson 2012, Melchert To appear), con�rmed by the forms
in section 3 and in general by media tantum (Watkins 1969: 113), Holli�eld 1977: 128, Vil-
lanueva Svensson 2003: 145, Villanueva Svensson 2012: 341).
Conclusion:
� (proto-)middle morphology in PIE occurred in both transitive and intransitive structures
� ... but some (PIE) transitive verbs lack synchronic motivation for middle morphology →�voice mismatch�, deponents
� PIE full grade middles (deponents or otherwise) do not imply an alternating paradigm (i.e.,
actives with a �higher� ablaut grade)
would solve both the problem of the missing digamma in the anlaut and the /˘u/ in the auslaut (these two properties correlateaccording to Hackstein 2002: 124f.), it requires giving up the equation with the Vedic forms, which lack initial s- and aremoreover anit
˙(cp. vr
˙tá-, vártar -, etc., EWA II: 512f.). On the other hand, there is no convincing alternative to the structural
analysis of Ved. varu° as *var-u-H ° (Klingenschmitt 1982: 233's proposal of /u/ as due to rhythmic lengthening is completelyad hoc), and the same holds for Greek ῥυ°. That is, there seems to be no way around the laryngeal metathesis proposed byHackstein.
11
Appendix
A. Reanalysis: self-benefactive/canonical middle > deponent
� Non-active morphology is a Spell-Out property of the functional projection v , active = default
(�elsewhere�) morphology:
(19) v ↔ v -X/_ No external argument (Embick 2004a: 150)
�Non-active voice is assigned when v does not introduce an external argument�
Arguments introduced below v (experiencers, benefactive arguments) trigger non-active morphol-
ogy. If such an argument is reanalyzed as agent, the result is a surface mismatch (agents are usually
in the speci�er of vP (Kratzer 1996) where they do not trigger non-active morphology). (20-a-b)
illustrate this reanalysis, the boxed parts are the arguments that undergo the reanalysis.
(20) a. vP
v(nonact)
ApplPben
DPbenApplben
Applben VP
V RP
R DP
b. vP
v [ag](nonact)
VP
DPagentV
V RP
R DP
B. Development of the *h2e-conjugation/proto-middle
1. Proto-middle
-h2e, *-th2e, *-e
2. Middle
-h2e(r), *-th2e(r), *-o(r)
4. �Dentalless middle�
-h2e(r/i“), *-th2e(r/i
“)/*-soi
“, *-o(r/i
“)
IIr. �stative�
Hitt. a(ri)-middles
OIr. berar -passive
5. PIE middle
-(m)h2e(r/i“), *-th2e(r/i
“)/*-soi
“, *-to(r/i
“)
3. �Neo-active�
PIE perfect
thematic conjugation
Hitt. hˇi -conjugation
12
References
Alexiadou, Artemis. 2013. Where is non-active morphology? In Proceedings of the 20th International Conference
on Head-Driven Phrase Structure Grammar , ed. S. Müller, 244�62. CSLI publications.
Alexiadou, Artemis, and Edit Doron. 2012. The syntactic construction of two non-active voices: passive and
middle. Journal of Linguistics 48:1�34.
Benveniste, Émile. 1969. Le vocabulaire des institutions indo-européennes, vol. II: pouvoir, droit, religion. Paris:
Éditions de Minuit.
Buck, Carl Darling, and Walter Petersen. 1945. A Reverse Index of Greek Nouns and Adjectives: Arranged by
Terminations with Brief Historical Introductions. Chicago: University of Chicago Press.
Embick, David. 1998. Voice systems and the syntax/morphology interface. In Papers from the UPenn/MIT
Roundtable on Argument Structure and Aspect , ed. H. Harley, 41�72. MIT Working Papers in Linguistics 32.
Embick, David. 2004a. Unaccusative syntax and verbal alternations. In The Unaccusativity Puzzle, ed. A. Alex-
iadou, E. Anagnostopoulou, and M. Everaert, 137�58. Oxford University Press.
EWA II = Manfred Mayrhofer. 1996. Etymologisches Wörterbuch des Altindoarischen, volume II. Heidelberg:
Universitätsverlag Winter.
Feist, Sigmund. 1939. Vergleichendes Wörterbuch der gotischen Sprache. Mit Einschluss des Krimgotischen und
sonstiger zerstreuter Überreste des Gotischen. Dritte neubearbeitete und vermehrte Au�age. Leiden: Brill.
Fortson, Benjamin W. IV. 2010. Indo-European Language and Culture: an Introduction. Blackwell, 2 edition.
García Ramón, José Luis. 1993. Griego ζατεω (: hom.-jon. διζημαι), véd. ya 2 �pedir� e IE *i“eh2- �pedir, buscar
afanosamente�. In Miscellanea Linguistica Graeco-Latina, ed. L. Isebaert, 71�84. Namur: Société des études
classiques.
García Ramón, José Luis. 1999b. Griechisch Ζητηρ · Ζευ`ς ἐν Κυπρῳ, vedisch yatár - �Rächer� und die Vertretung
von *i“- im Griechischen. In Compositiones Indogermanicae in Memoriam Jochem Schindler , ed. H. Eichner,
H.C. Luschützky, and V. Sadovski, 77�96. Prague: Enigma.
Geldner, Karl F. 1951. Der Rig-Veda aus dem Sanskrit ins Deutsche übersetzt, vol. I-III . Harvard Oriental Series
33-35. Cambridge, Mass.
Geniu²iene, Emma. 1987. The typology of re�exives. Berlin/New York/Amsterdam: De Gruyter.
Gonda, Jan. 1960. Re�ections on the Indo-European medium. Lingua 9:30�67; 175�93.
Grestenberger, Laura. 2014. Feature Mismatch: Deponency in Indo-European. Doctoral Dissertation, Harvard
University.
Hackstein, Olav. 1995. Untersuchungen zu den sigmatischen Präsensstammbildungen des Tocharischen. Göttin-
gen: Vandenhoeck & Ruprecht.
Hackstein, Olav. 2002. Die Sprachform der homerischen Epen. Wiesbaden: Reichert.
Hintze, Almut. 2000. Die avestische Wurzel mad `zumessen'. In Indoarisch, Iranisch und die Indogermanistik.
Arbeitstagung der Indogermanischen Gesellschaft vom 2.-5. Oktober in Erlangen, ed. B. Forssman and R. Plath,
163�75. Wiesbaden: Reichert.
Holli�eld, Patrick. 1977. On the System of Conjugation in Proto-Indo-European. Doctoral Dissertation, Harvard
University.
Insler, Stanley. 1968. The origin of the Sanskrit passive aorist. Indogermanische Forschungen 73:312�46.
Jamison, Stephanie W., and Joel P. Brereton. 2014. The Rigveda: The Earliest Religious Poetry of India, vol.
I-III . Oxford University Press.
Jasano�, Jay. 1988. PIE *“gne- `recognize, know'. In Die Laryngaltheorie und die Rekonstruktion des indoger-
manischen Laut- und Formensystems, ed. A. Bammesberger. Heidelberg: Winter.
Jasano�, Jay. 2003. Hittite and the Indo-European Verb. Oxford University Press.
13
Jasano�, Jay. 2012a. Did Hittite have si -imperatives? In The Sound of Indo-European II , ed. R. Suka£ and
O. �ef£ík. München: Lincom.
Jasano�, Jay. 2012c. Long-vowel preterites in Indo-European. In The Indo-European Verb. Proceedings of the
Conference of the Society for Indo-European Studies, Los Angeles 13-15 Sept. 2010 , ed. H. C. Melchert, 127�35.
Wiesbaden: Reichert.
Kallulli, Dalina. 2007. Rethinking the passive/anticausative distinction. Linguistic Inquiry 38/4:770�80.
Kallulli, Dalina. 2013. (Non-)canonical passives and re�exives: deponents and their like. In Non-Canonical
Passives, ed. A. Alexiadou and F. Schäfer, 337�58. Amsterdam/Philadelphia: John Benjamins.
Kaufmann, Ingrid. 2007. Middle voice. Lingua 117:1677�714.
Kemmer, Suzanne. 1993. The Middle Voice. Amsterdam: John Benjamins.
Kemmer, Suzanne. 1994. Middle voice, transitivity, and the elaboration of events. In Voice: Form and Function,
ed. B. Fox and P. Hopper, 179�230. Amsterdam: John Benjamins.
Klaiman, Miriam H. 1991. Grammatical Voice. Cambridge University Press.
Klingenschmitt, Gert. 1982. Das altarmenische Verbum. Wiesbaden: Reichert.
Kloekhorst, Alwin. 2008. Etymological Dictionary of the Hittite Inherited Lexicon. Leiden: Brill.
Kratzer, Angelika. 1996. Severing the external argument from its verb. In Phrase Structure and the Lexicon, ed.
J. Rooryck and L. Zaring, 109�37. Dordrecht: Kluwer.
Kümmel, Martin. 1998. Wurzelpräsens neben Wurzelaorist im Indogermanischen. Historische Sprachforschung
111:191�208.
LIV2 = Rix, Helmut. 2001. Lexikon der Indogermanischen Verben. Bearbeitet von Martin Kümmel, Thomas
Zehnder, Reiner Lipp und Brigitte Schirmer . Wiesbaden: Reichert, 2nd edition.
Malzahn, Melanie. 2010. The Tocharian Verbal System. Leiden: Brill.
Masson, Olivier. 1983. Les inscriptions cypriotes syllabiques. Paris: École francaise d'Athènes.
Mayrhofer, Manfred. 1986. Indogermanische Grammatik I, 2. Halbband: Lautlehre [Segmentale Phonologie des
Indogermanischen] . Heidelberg: Winter.
Melchert, H. Craig. 1994. Anatolian Historical Phonology . Amsterdam: Rodopi.
Melchert, H. Craig. 2011. The PIE verb for `to pour' and medial *h3 in Anatolian. In Proceedings of the 22nd
UCLA Indo-European Conference, Los Angeles Nov. 5-6 2010 , ed. S. Jamison, H. C. Melchert, and B. Vine,
127�32.
Melchert, H. Craig. To appear. �Narten formations� versus �Narten roots�. In Proceedings of the Abeitsta-
gung of the Society for Indo-European Studies on the Lengthened Grade in Indo-European. Indogermanische
Forschungen.
Narten, Johanna. 1964. Die sigmatischen Aoriste im Veda. Wiesbaden: Reichert.
Pinault, Georges-Jean. 1982. L'expression indo-européenne de la nomination. Études indo-européennes 3:15�36.
Rau, Jeremy. 2009. Indo-European Nominal Morphology: the Decads and the Caland System. Innsbrucker
Beiträge zur Sprachwissenschaft 132. Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck.
Rau, Jeremy. 2013. Notes on state-oriented verbal roots, the Caland system, and primary verb morphology in
Indo-Iranian and Indo-european. In Multi Nominis Grammaticus: Studies in Classical and Indo-European
Linguistics in Honor of Alan J. Nussbaum on the Occasion of his Sixty-�th Birthday , ed. A.I. Cooper, J. Rau,
and M. Weiss, 255�273. Ann Arbor/New York: Beech Stave.
Solmsen, Felix. 1901. Untersuchungen zur griechischen Laut- und Verslehre. Strassburg: Trübner.
Villanueva Svensson, Miguel. 2003. La categoría de voz en el sistema verbal indoeuropeo. Doctoral Dissertation,
Universidad complutense de Madrid.
Villanueva Svensson, Miguel. 2006. Old Lithuanian pame(d)mi `imitate'. Baltistica 41/1:95�7.
Villanueva Svensson, Miguel. 2012. The ablaut of the middle root athematic presents in Indo-European. In The
14
Indo-European Verb. Proceedings of the Conference of the Society for Indo-European Studies, Los Angeles,
13-15 September 2010 , ed. H. C. Melchert, 333�42. Wiesbaden: Reichert.
Watkins, Calvert. 1962. Varia. II. Ériu 19:114�8.
Watkins, Calvert. 1969. Geschichte der indogermanischen Verbal�exion. Indogermanische Grammatik, III,1.
Heidelberg: Winter.
Weiss, Michael. 2009. Outline of the Historical and Comparative Grammar of Latin. Beech Stave Press.
Zombolou, Katerina, and Artemis Alexiadou. 2014. The canonical function of the deponent verbs in Modern
Greek. In Morphology and Meaning. Selected Papers from the 15th International Morphology Meeting, Vienna,
February 2012 , ed. F. Rainer, F. Gardani, H. C. Luschützky, and W. U. Dressler, 331�44.
15